//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
278 [278–300] 31.7.2008 11:10AM
CHAPTER FIFTEEN
Temporal patterns in diversification
rates
andy purvis, c. david l. orme, nicola h.
toomey and paul n. pearson
Introduction
The study of rates of speciation and extinction, and how these have changed
over time, has traditionally mainly been the preserve of paleontology (Simpson
1953; Stanley 1979; Raup 1985). More recently, phylogenies of extant species
have been shown to contain information on these rates and how they may have
changed, under the assumption that the same rules have applied in all contemporaneous lineages (Harvey et al. 1994; Kubo & Iwasa 1995; see Nee 2006 for a
recent review). The first section of this chapter contrasts the strengths and
weaknesses of these two approaches – paleontological and phylogenetic – to
the study of macroevolution in general.
Moving to a specific macroevolutionary hypothesis, we then outline some tests
of the hypothesis that diversification rates have declined in the recent past, either
in response to changed abiotic conditions or as a result of density-dependence
or diversity-dependence. It has long been appreciated that incomplete specieslevel sampling can cause a bias in favour of this hypothesis at the expense of the
null hypothesis of no change (Pybus & Harvey 2000), but we highlight a further
sort of incompleteness that is likely to be very widespread and which is not
widely appreciated – products of recent lineage splits are unlikely to be considered as distinct species. We reanalyze the data from a key early paper (Zink &
Slowinski 1995) to show how this incompleteness, which is inevitable when
taxonomy and phylogeny meet, is sufficiently strong to account for much
(though not all) of the apparent tendency for rates to have declined through time.
Attempts to infer the deeper history of diversification rates are relatively free
of this problem but instead encounter others: in particular, the assumption
that the underlying probabilities of speciation and extinction per unit time
have changed in the same way in all lineages becomes increasingly unlikely as
a wider and wider clade is considered. We explain how analyses of a nearcomplete species-level phylogeny of extant mammals show a complex pattern
of temporal variation in the rate of effective cladogenesis (Bininda-Emonds et al.
2007), and compare the mammalian picture with that obtained some years
previously for birds (Nee et al. 1992).
Speciation and Patterns of Diversity, ed. Roger K. Butlin, Jon R. Bridle and Dolph Schulter. Published by
Cambridge University Press. # British Ecological Society 2009.
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
279 [278–300] 31.7.2008 11:10AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Phylogenies of extant species do not contain information to decide whether
such temporal variations reflect changes in speciation rates, extinction rates or
both. This information can only ever be obtained from the relatively few fossil
records that are sufficiently well sampled to permit sophisticated analysis at the
species level. The final section of the chapter shows how a model system
approach to macroevolution would inform and develop the field as a whole.
We propose Tertiary planktonic foraminifera as the most promising model
system, and show how such systems have the potential to provide insights
into how clades wax and wane that is simply not available from data on extant
species alone.
Two approaches to macroevolution
Figure 15.1a is a cartoon of the ideal data set for macroevolutionary research.
The fossil record of this obliging clade is comprehensive and continuous, permitting lineages to be traced accurately through the rocks: lineage splits are
speciations, and the sampling is so complete that the disappearance of a lineage
marks its extinction. This level of detail greatly facilitates identification of nonrandom patterns in the clade’s history. For example, four of the lineages in
Figure 15.1a went extinct within a very narrow time window; and the clade
comprising the two extant species on the left has had both lower speciation rate
and a lower extinction rate than the clade comprising the remaining extant
species. Furthermore, the ideal data set also contains the history of many species
attributes: morphometric features such as body size and shape can be read
directly at any point in time, as can each species’ geographical range, paleotemperature and so on. Provided that such a data set is large enough to give
reasonable statistical power, hypothesis-testing is easy: for instance, rates
(whether of speciation, extinction or anagenetic change) can be correlated
with attributes of both species and their environments (Foote 1996; Jablonski &
Roy 2003); character evolution can be traced simply along lineages to test
hypotheses about trends or stable attractors (Alroy 2000).
Unfortunately, this ideal is never reached. Paleontologists typically have a
data set that is more like the caricature in Fig. 15.1b, so testing hypotheses
reliably is more difficult. More – usually much more – of the fossil record is
missing than present, and what is present is not fully representative. Some
lineages (e.g. the two right-hand species) have no fossil record at all, and the
same is true of many time periods (some short, some long); sample completeness is likely to vary over time in a complicated fashion that can be hard to
correct for (Alroy et al. 2001; Smith et al. 2001; Peters & Foote 2002). Occasionally
a lineage can be traced for a long time, but the record for any lineage is usually
very fragmentary. Consequently, speciation events and extinction events are
much harder to infer correctly: many will be missed, because the species is
entirely absent from the record, and others may be incorrectly inferred to have
279
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
Time
PURVIS ET AL.
Species attributes
Time
280
280 [278–300] 31.7.2008 11:10AM
Species attributes
Figure 15.1 Caricatures of data sets for macroevolution. (a) The ideal data set; speciations
(diamonds), extinctions (crosses), relationships and attributes are all directly available
throughout the group’s history. (b) A paleontological data set. The record is too
fragmentary to confidently infer dates of speciation or extinction, and has gaps that are
non-random with respect to both lineage and time. (c) A data set restricted to present-day
diversity. Relationships can be inferred (straight lines), and the history of attributes can
be inferred (position along attribute axis), but neither extinctions nor attribute history
are recoverable directly.
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
281 [278–300] 31.7.2008 11:10AM
Time
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Species attributes
Figure 15.1 (cont)
taken place. Such pseudoextinction and pseudospeciation arise when a single
evolving lineage of ancestor–descendant populations is split more or less
arbitrarily into multiple morphospecies, the type specimens of which may be
very distinct (Stanley 1979; Pearson 1998a). Analyses of evolutionary dynamics
that are based on morphospecies therefore risk conflating rates of speciation
with rates of anagenetic change (Pearson 1998b). It can often be difficult to crossreference timescales among geographic regions, adding uncertainty to the timing of inferred events. Additionally, the attribute data are likely to be even more
incomplete than the figure shows: most groups have geographically patchy
coverage, for instance, complicating inferences relating biogeography to clade
dynamics.
The phylogenetic approach has become extremely popular in recent years,
with the explosion in molecular systematics making it very much easier to
produce more or less complete species-level phylogenies than more or less
complete species-level fossil records. The sophistication of phylogeny estimation procedures has increased greatly in recent years, though there is still
ongoing development of ‘relaxed clock’ methods for obtaining relative dates
of branching points, and of procedures for using the fossil record to calibrate
these to absolute time (Benton & Donoghue 2007). Attributes of present-day
species – including those that do not fossilize – can also be measured relatively
easily and directly. However, even if the data and phylogeny were absolutely
perfect, the neontologist’s view of a clade’s history is still woefully incomplete
281
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
282
282 [278–300] 31.7.2008 11:10AM
PURVIS ET AL.
(Fig. 15.1c). One obvious problem, though not our main focus in this chapter, is
that the attributes of ancestors have not been observed, and so must be inferred
from the attributes of their extant descendants, using the phylogeny and some
model of how attributes change through time. It is hard to know what model to
use, and estimates of ancestral characteristics will often be unreliable (see, e.g.
Cunningham et al. 1998; Oakley & Cunningham 2000; Webster & Purvis 2002).
Another serious problem is that extinctions are completely missing from the
picture in Figure 15.1c. Nodes in the phylogeny represent speciations, but only
those speciations from which more than one daughter branch has an extant
descendant: other speciations are missed. This obviously gives a limited view of
what has happened, and has also led to awkward terminology: the rate at which
nodes occur on the tree is not the rate of speciation, nor of cladogenesis, so is
typically termed either the net rate of diversification (Zink & Slowinski 1995) or
the rate of effective cladogenesis (Nee et al. 1992).
Faced with an absence of direct information about the past state of the system
under study, researchers typically must resort to simple models of what might
have happened. When the issue is diversification rates through time, the simple
models are likely to be pure birth or birth–death models, in which all species, or
all contemporaneous species, have the same chances of speciation and, in a
birth–death process, extinction. When the issue is character evolution, the
model is likely to be a continuous-time Markov model (Felsenstein 1985; Pagel
1994; Schluter et al. 1997). These models are either fitted, to produce parameter
estimates, or used as null models to test whether some marginally more complex alternative provides a better fit to the data. The next two sections consider
how such approaches have been used to test for temporal patterns in the
diversification rate of clades.
Tests for simple changes in the net rate of diversification
Extinctions are missing from the data in Fig. 15.1c, and are also missing from
one of the models most commonly used in macroevolutionary analyses of
molecular phylogenies. In the pure birth process, or Yule process (after Yule
1924, who first developed the model mathematically for the study of macroevolution), each species has the same constant instantaneous speciation rate, λ,
and there is no extinction. Cladogenesis is modelled as a stochastic process: the
waiting time before a given species speciates has a mean of 1/λ, but is exponentially distributed. This simple model leads to many straightforward but powerful results (see Nee 2001, 2006). Two are of particular interest here. One is that a
clade should grow exponentially, such that a plot of ln(lineage number) against
time should produce a straight line with slope λ. Such lineages-through-time
plots (Nee et al. 1992) have been widely used to give a first impression of clade
dynamics. A closely related result concerns the timing between successive
speciation events as a clade diversifies under this process. The expected wait
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
283 [278–300] 31.7.2008 11:10AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
before any of nt species in a clade speciates is 1/nt λ. So successive speciation
events are separated by ever less time. However, with nt species, the expected
amount by which the phylogeny’s total branch length increases during the wait
for the next speciation is nt/nt λ = 1/λ (Purvis et al. 1995). Under a Yule process,
then, the phylogeny of a clade grows by (on average) the same amount between
each pair of speciation events, and this amount is simply the reciprocal of the
per-lineage speciation rate.
These two results provide baselines against which to judge the pattern of node
ages in molecular phylogenies. If λ has not been constant, but has been increasing through time, then the lineages-through-time plot will steepen towards the
present, and successive speciations will have been accompanied by less and less
growth in the total branch length of the phylogeny. Conversely, if λ has been
decreasing with time, the lineages-through-time plot will tend to flatten, and
the phylogeny will on average grow by more and more between successive
speciation events.
Extinction changes the picture. At least, it should. However, a common way
that researchers have tried to generalize the Yule process to include extinction
makes no difference. If speciation rate is λ and extinction rate is μ, then it is
tempting to model the resulting birth–death process as a pure birth process with
speciation rate λ − μ, and statistical tests have been developed that use this
model to test whether diversification rates have changed through time (Paradis
1997, 1998; Price et al. 1998). However, the birth–death process cannot be
modelled in this way (Nee 1994). If μ > 0 then, even if λ and μ are constant through
time, the lineages-through-time plot steepens towards the present: the expected
slope for much of a clade’s history is indeed λ – μ but, starting around 1/(λ – μ) time
units ago, it starts to steepen until, at the present day, the slope estimates λ (Nee
et al. 1994). Equally, the phylogeny is expected to grow by ever smaller amounts
between successive speciation events. Consequently, most tests are unable to
discriminate between a true increase in the net rate of diversification and a
constant-rates birth–death process (Pybus & Harvey 2000).
Decreases in the net rate of diversification are another matter, however. They
are not expected from a complete species-level phylogeny that has grown under
a birth–death process. They are also fascinating theoretically, because they are
consistent with density-dependent diversification, which some (though not all)
large-scale paleontological analyses support (see, for examples, Alroy 1996,
1998; Sepkoski 1996; Foote 2000; Kirchner & Weil 2000).
In a key paper, Zink and Slowinski (1995) found that small (N = 3 to 11 species)
but mostly complete species-level molecular phylogenies of 11 North American
bird genera mostly showed diversification rate to have been decreasing rather
than increasing. Ten genera were large enough to assess the direction of curvature in the lineages-through-time plot, and nine of these (significantly more
than half) showed a decrease rather than an upturn. Zink and Slowinski also
283
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
284
284 [278–300] 31.7.2008 11:10AM
PURVIS ET AL.
developed a test statistic that uses the branch length increases between successive speciations, and which was intended to follow a standard Normal distribution under a pure birth process; 10 of the 11 genera (again, significantly more
than half) had a negative value of this statistic. Some individual genera showed
significant slowdown (though one-tailed tests were used incorrectly, so the pvalues should all be doubled), but the startling feature of the analysis was that
such a high proportion of the data sets showed a tendency towards slowdown.
Pybus and Harvey (2000) corrected an error in Zink and Slowinski’s test
statistic, calling the corrected version γ. They explicitly drew attention to how
incomplete sampling biases γ downwards (that is, towards a result that supports
slowdown) and proposed using Monte Carlo simulations to test whether an
apparent slowdown remained significant when the (known) level of sample
incompleteness was accounted for, under an assumption that species are missing from the phylogeny at random. They noted that overdispersed and underdispersed samples would bias γ downwards and upwards, respectively. This
method has been widely used since, with results often showing significant
slowdown of diversification (see Price & Phillimore, this volume, for a review).
There is one important way in which sampling is often likely to be nonrandomly incomplete. Because speciation is usually a gradual process, there
are unlikely to be many pairs of sister species that split in the very recent past.
Such recent lineage splits are unlikely to be deemed speciation events: a node is
likely to be designated as a speciation event only if both lineages persist long
enough to evolve differences that attract taxonomic attention.
In order to illustrate how species designation depends on node age, we have
used published data on the levels of cyt b sequence divergence for splits that were
within species (Avise & Walker 1998; Avise et al. 1998) or between putatively sister
species (Johns & Avise 1998). Due to sample incompleteness, some of these latter
nodes probably delimit clades of more than two species, and the estimated
‘sister’-species sequence divergence does indeed decrease linearly as the proportion of recognized species sampled with each genus increases. We have used
residuals around this relationship, converted to units of time using the 2% per
million years calibration, as corrected estimates of sister-species divergence. As
the intraspecific divergence times presented in Avise and Walker (1998), and
Avise et al. (1998) represent the upper limit of possible between-population
divergence, we have used the reported numbers of populations for each study
as additional estimates of intraspecific divergence. The actual divergence estimates for these between-population splits are not reported, so we have made the
assumption that these splits have zero sequence divergence; this assumption is
conservative with respect to the point we wish to make here. Divergences were
converted to units of time as in the papers from which they came. The full data set
was then used in a binomial general linear model (logistic regression) to predict
whether divergence time and vertebrate group (mammal, bird, herpetofauna or
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
285 [278–300] 31.7.2008 11:10AM
Interspecific
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Intraspecific
Birds & Fish
Herpetofauna & Mammals
10
8
6
4
2
0
Divergence time (Mya)
Figure 15.2 Plot showing effect of estimated divergence time on whether a branching
event is deemed a speciation event. Curves show predictions from a binary general linear
model including vertebrate group (fish, herpetofauna, mammal or bird) as an additional
predictor; model simplification indicates birds and fish share a model, as do mammals
and herpetofauna.
fish) predict whether a given divergence represents a speciation event, with
model simplification used as appropriate (Crawley 2002). The resulting model is
shown in Fig. 15.2. In each vertebrate group, the data suggest that recent splits are
very unlikely to be designated as speciation events, especially in birds and fish.
Other data sets would doubtless give quantitatively different results, but the
qualitative pattern is an inevitable consequence of grafting species-level taxonomy and lineage phylogeny together.
Figure 15.3 shows the implication of the pattern for studies of slowdown in
species-level phylogenies. If the truth has been a Yule process, but recent splits
are unlikely to be included in the phylogeny (which is nonetheless complete,
inasmuch as it contains all the recognized species in a group), then the lineagesthrough-time plot or γ are biased downwards (that is, towards slowdown).
To what extent could this bias be responsible for the prevalence of negative γ
values in published studies? We combined simulations of a Yule process with the
above species designation model to produce a preliminary assessment. The value
of λ came from pooling an illustrative set of 14 avian species-level phylogenies
(incorporating those of Zink and Slowinski that we could replicate) to find the
average λ from early in a clade’s history. We used the 14 series of internode
distances (as percentage sequence divergence) and calculated the mean of each
successive internode distance from the root across the groups (i.e. averaging all
the first internodes, then all the second internodes, and so on) to obtain a
285
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
PURVIS ET AL.
Number of lineages
16
Recognition probability
(a)
1.0
(b)
8
4
2
(c)
0.8
0.6
0.4
0.2
0.0
(d)
Number of lineages
286
286 [278–300] 31.7.2008 11:10AM
16
(e)
8
4
2
6
5
4
3
2
Time before present (Mya)
1
0
Figure 15.3 Cartoon illustrating how
AQ1
apparent slowdown in diversification
can arise simply through age-dependency
in whether nodes are deemed to be
speciation events. A phylogeny that has
grown under a constant rate, pure birth
model (a) yields a lineages-through-time
plot with a constant slope (b). If the
probability that cladogenetic events
within the phylogeny are recognized as
speciation events varies as a function of
the age of the event (c), then the resulting
species phylogeny (d; dotted branches are
not viewed as separate species) will tend
to yield lineages-through-time plots that
exhibit apparent decrease in net
diversification (e).
lineages-through-time plot averaged across the 14 groups. Our estimate of early λ
was then the slope of the first segment in a breakpoint regression of ln(N) against
time (constrained to have an initial intercept of ln(2), and with the breakpoint
chosen by least-squares). One thousand Yule phylogenies were then simulated
with this value of λ, with crown groups the same age as the average among our
14 clades. We computed γ for each of these 1000 simulated trees.
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
287 [278–300] 31.7.2008 11:10AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Table 15.1 Slowdown in 14 avian phylogenies, showing γ and p for each group
when analysed at face value, along with the number of species (n) and the sources
Taxon
γ
p
n
Source
Ammodramus†
Anasa
Dendroicaa
Diomedeidae
Gruidae
Melospiza†
Parus (Chickadees)†
Parus (Titmice)†
Passerella†
Pipilo†
Quiscalus†
Spizella†b
Toxostoma†
Zonotrichia†
− 1.449
− 0.296
− 4.173
− 0.509
− 2.413
− 1.410
− 2.645
0.288
− 1.633
− 1.255
− 1.182
− 0.924
− 2.490
0.486
0.074
0.383
<0.001
0.305
0.008
0.079
0.004
0.613
0.051
0.105
0.119
0.178
0.006
0.687
8
27
26
14
15
3
7
4
4
4
4
6
11
6
Zink and Avise (1990)
Johnson and Sorenson (1998)
Lovette and Bermingham (1999)
Nunn et al. (1996)
Krajewski and Fetzner (1994)
Kessler and Avise (1985)
Gill et al. (1993)
Gill and Slikas (1992)
Zink (1994)
Zink and Dittman (1991)
Zink et al. (1991b)
Zink and Dittman (1993)
Zink et al. (1999)
Zink et al. (1991a)
Note: Phylogenies used by Zink and Slowinski (1995) are indicated (†). The phylogenies
contain all described species except where marked: (a) >90% described species; (b) <90%
described species.
To simulate the effect of age-biased sampling, we used the species designation
model to give the probability that nodes of a given age would be sampled from
the simulated trees: older nodes within a tree are more likely to be sampled than
younger nodes, as in Fig. 15.2. Unsampled nodes were pruned from the trees,
and each tree’s γ was computed. By comparing the values of γ from before and
after pruning, we were able to estimate the likely effect of age-dependent
sample incompleteness on the γ values in the original analyses of these phylogenies. Before pruning, the mean γ was − 0.19 (significantly negative, as pointed
out by Price & Phillimore, this volume); pruning reduces the mean to − 0.75.
Table 15.1 shows the γ for each of the 14 phylogenies; the centre of the distribution is significantly less than zero (Wilcoxon signed ranks test: p = 0.0004), but
only marginally significantly less than − 0.75, the mean for the pruned simulations
(Wilcoxon signed-ranks test: p = 0.045). This result suggests that the age-dependent
probability of designating a node as a speciation event could be responsible for
much of the apparent slowdown in diversification in these avian phylogenies, and
particularly for much of the general tendency for slowdown.
One of the phylogenies does show strong evidence for slowdown: Dendroica, with
24 species, is the largest of the set, and is large enough to attempt to fit the same
sort of mode to it on its own. However, the initial phase of diversification in this
genus is so rapid that Yule processes with the estimated λ grow too large for our
287
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
288
288 [278–300] 31.7.2008 11:10AM
PURVIS ET AL.
software before they reach the present day, implying that the slowdown is much
more severe than age-dependent incompleteness could produce (Orme 2002).
If age-dependent sample incompleteness biases tests of slowdown, what can
be done about it? Increasingly, sequence data are available for many populations within species, which may permit a solution. Pons et al. (2006) show how
the difference in expected branching pattern between within- and betweenspecies phylogenies can be used to identify the depth in the tree that best
represents the species level. For groups with sufficiently rich data sets, it is
therefore possible to prune the phylogeny back to this depth, and used the
pruned phylogeny as the basis for testing.
Tests for more complex temporal patterns
The γ test reduces the information available in a molecular phylogeny to a single
number, which provides insight into whether the net rate of diversification
decreased or increased over time. In reality, clade dynamics may have shown a
more complex temporal pattern of diversification, so there is a need for an
approach that can be used to test more complex hypotheses. Two sorts of
approaches have been developed recently. The first views a clade’s history as
having had multiple (usually two) different birth–death regimes, with a sudden
transition between them. The timing of the transition(s), and the different
speciation and extinction rates, are estimated from the timings of nodes in the
phylogeny, and a range of approaches are available to test whether the additional complexity is merited by the data (Barraclough & Vogler 2002; Turgeon
et al. 2005; Rabosky 2006).
Bininda-Emonds et al. (2007) developed a second approach, using generalized
additive models (GAMs; Wood 2006) to model the net rate of diversification as a
smooth function of time – a curve rather than a set of steps. Unlike standard
regression approaches, GAMs do not require a detailed specification of the
nature of the relationship being modelled. This feature is particularly attractive
here, because there is generally no a priori reason for preferring a particular
form. The GAM provides a model with minimum complexity necessary to
capture the relationship, and also provides the facility for testing whether the
relationship is consistent among clades (as an ANCOVA would do in a linear
model framework). Their analysis took its data from a near-complete and dated
species-level composite ‘supertree’ phylogeny of extant mammals (BinindaEmonds et al. 2007). Although the phylogeny is almost complete, it is only
around 46% as resolved as a fully bifurcating tree would be, with most of the
lack of resolution being near the tips (especially within genera; the phylogeny
prior to 32 MYA is 75% resolved) and within a few poorly studied groups (notably
Muridae). Also, the dates of around one third of the nodes in the tree are not
estimated directly from sequence data or fossils, but interpolated based on the
dates of surrounding nodes and on the diversities of the clades involved.
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
289 [278–300] 31.7.2008 11:10AM
Miocene
Oligocene
Eocene
0.00
0.05
Diversification rate
0.10
0.15
Paleocene
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
150
100
50
10
Time before present (My)
Figure 15.4 Mammalian net diversification rate through time. The set of grey steps
represent the average rates within subepochs. The solid curve is the fitted rate from a
generalized additive model (GAM); the dashed lines are 95% confidence intervals. The
thick upright grey line at 65.5 Mya represents the Cretaceous-Tertiary transition; other
vertical lines demarcate Tertiary epochs. After Bininda-Emonds et al. (2007).
Consequently, it would be unwise to use all the branches in the phylogeny at
face value. The analysis was therefore restricted to only those branches whose
start and end dates were known most reliably, that is, branches that did not start
at a polytomy, and whose start and end dates were estimated from data rather
than interpolated. Over 3100 branches, many of them ending in a present-day
species, met these criteria. These are not a random sample of the branches in the
tree – resolution is poorest within genera – which leads to predictable biases in
the results; we describe and discuss these below.
The net rate of diversification was first estimated in each geological subepoch
using a survival analysis (where a node, counter-intuitively, marks the ‘failure’
of the parent lineage), as shown in Fig. 15.4. This analysis made clear that the
temporal pattern of net diversification rate was indeed likely to be complex
(stepped line in Fig. 15.4). The analysis then moved to a 0.1 My timescale,
because this was the level of resolution in the dates in the phylogeny, and all
the branches in the analysis are at least this long. Within each 0.1 My interval,
the lineages present that did and did not result in a node at the end of the
interval were counted, and turned into a binomial response variable. The
advantage of this approach over simply computing a rate is that the binomial
response variable contains sample size information, so the subsequent model is
289
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
290
290 [278–300] 31.7.2008 11:11AM
PURVIS ET AL.
appropriately weighted. The response variable was then modelled as a smooth
function of time. Because the relationship could in principle be very complex,
with peaks or troughs lasting only one or a few million years (very short
compared to the full 166 My time scale), the basis dimension (k, a parameter
that sets the maximum complexity of the curve) for the fitted curves was set
high enough to permit a kink in the curve every 1 My; using the default value of k
permitted only ten kinks, which made the curve noticeably less responsive to
the data. Further details of the analysis can be found in Bininda-Emonds et al.
(2007). The curved lines in Fig. 15.4 show the fitted rates and confidence intervals plotted against time.
The right-hand side of the curve is hard to interpret for three reasons. First,
the fact that recent lineages are unlikely to be designated as species is at least
partly responsible for the very low rates of diversification in the very recent past.
Second, this low rate is likely to also be partly due to a tendency for only speciespoor genera to have well-resolved phylogenies: branches that are in reality long
are more likely to meet the criteria for inclusion in the analysis. This bias will be
much less marked deeper in the tree, but is likely to suppress estimates of the
net diversification rate for the last few million years. Third, some of the preceding increase in rate will be the upturn expected under a birth–death process.
This artifact has the longest reach into the past, but is expected to subside
within about 1/(λ – μ) time units ago (Harvey et al. 1994); this is basically because,
if a clade survives for 1/(λ – μ) and l > m, it will very probably have attained a high
enough diversity to make future stochastic extinction unlikely. Bininda-Emonds
et al. (2007) therefore restricted their interpretation to the period earlier than
25 My ago (the reciprocal of their lowest net rate of diversification), focusing on
the peak in rates from 100–85 My ago, the decline in rates that persisted through
the end-Cretaceous mass extinction, and then a subsequent upturn that did not
start until around the early Eocene, about 55 My ago. Their analysis showed no
indication of any significant pulse or other increase in the rate of origin of
extant mammalian lineages after the end-Cretaceous event, except for a possible increase in marsupials: there, lineage number doubled around the
Cretaceous–Tertiary boundary, but the numbers of lineages involved are too
small (rising from 3–6) for any firm conclusion to be reached. Intriguingly, the
lack of an upturn in the earliest Tertiary agrees completely with a much earlier
analysis of Sibley and Ahlquist’s (1990) DNA–DNA hybridization phylogeny of
birds. Nee et al. (1992) analysed the portion of the tree for which lineage
sampling was probably complete; according to the time calibration used by
Sibley and Ahlquist, this was prior to 45 MY ago. They used a range of tests to
demonstrate that the per-lineage net rate of diversification had decreased
throughout the preceding history of bird diversification, with the lineagesthrough-time plot showing a similar shape to the one reported by BinindaEmonds et al. (2007). The phylogeny underpinning this analysis is controversial,
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
291 [278–300] 31.7.2008 11:11AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
but it will be interesting to see whether the similarity in temporal pattern
persists when more robust phylogenies are available for birds. It is also important to note that both sets of dates, along with many other timescales derived
from molecular phylogenies, are often considerably older than corresponding
estimates taken directly from the fossil record, which would have modern
orders originating after the Cretaceous–Tertiary boundary (Wible et al. 2007).
In dating the tree, Bininda-Emonds et al. (2007) took the view that a fossil
provides only a minimum estimate for the age of the crown-group clade within
which it is nested. This is because, with an incomplete fossil record, the early
days of the crown group are likely to have left no fossils that are both known and
correctly identified; the more incomplete the record, the longer such gaps will
tend to be (Tavaré et al. 2002). An alternative view is that the fossil record is
complete enough that at least some of the fossils used in calibration provide
more than a minimum estimate, indicating that a clade must have originated
only soon before the time those species lived. Reconciling these views is a
challenge for future research (Cifelli & Gordon 2007; Penny & Phillips 2007).
The GAM approach permits great flexibility in modelling. It is possible to test
for significant differences among clades, analogous to the use on ANCOVAs as an
extension of standard regressions: the two major superorders, Euarchontoglires
and Laurasiatheria, show broadly similar patterns, while the other two placental
superorders combined (Xenarthra and Afrotheria) show the early high rate but do
not show the rate increase at 55 MY ago that these two groups show (BinindaEmonds et al. 2007). It is also possible to use GAMs to test whether the response
variable shows a sharp disjunction at a time thought in advance to be important
(Wood 2006), such as ends of epochs. Further, it has the advantage that it can be
used even if the start and end dates are not known for all the branches in the
phylogeny, provided that it is reasonable to assume that the branches for which
these dates are available are a random or otherwise representative sample of the
whole phylogeny. Also, the focus on deeper branches in the phylogeny make it
easier in principle to meet the requirement for completeness, and make it possible to test for increases as well as decreases in the net rate of diversification.
Although these methods have some advantages over earlier ones developed to
answer similar questions, they are still inherently limited by their reliance on
data from present-day diversity. For example, there is excellent fossil evidence
from North America indicating that mammalian speciation rates spiked in the
immediate aftermath of the end-Cretaceous event (Alroy 1999, 2000), but the
phylogeny of present-day species contains no significant signal of this event,
because the great majority of the new species were in groups that have subsequently declined or gone extinct (Alroy 2000; Bininda-Emonds et al. 2007). In
this instance, the paleontological and present-day approaches give complementary insights. The North American fossil record reveals the speciation pulse, but
leaves open the possibility that still-extant lineages were diversifying in other,
291
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
292
292 [278–300] 31.7.2008 11:11AM
PURVIS ET AL.
less well-studied, parts of the world at the same time. The phylogeny contains
no evidence of the early Tertiary radiations that died out completely, and is
mute about whether speciation rate or extinction rate fluctuations have mattered more, but shows that the extant lineages were not diversifying in some
as-yet-unworked part of the world. One obvious priority for future macroevolutionary research is to develop statistical methods that more formally integrate
data from the fossil record and present-day diversity: both should reflect
the same underlying reality, and so frameworks combining these both should
lead to a more accurate and precise picture of what actually happened (Jablonski
et al. 2003).
Problems with the phylogenetic approach
Mammals have a good fossil record, compared with that of most groups, so it is
easy to see the ways in which data from present-day diversity alone give an
incomplete or even misleading picture of what happened. What of other
groups? The reliability of any inferences will depend upon the adequacy of the
models used to bridge the gap left by the lack of direct information about the
history of the system. Unfortunately, the data from present-day diversity do not
provide a good basis for testing whether the models are adequate. They are not
informative about whether, if net diversification rates have varied, variation is
driven by speciation or extinction rates (Barraclough & Nee 2001). They do not
directly record past diversity, weakening tests for density-dependence. They are
mute about whether individual characters have shown evolutionary trends
(Oakley & Cunningham 2000; Webster & Purvis 2002). These problems are
exacerbated when testing hypotheses in which diversification and character
evolution are linked: data from extant diversity alone may be unable to distinguish between very different scenarios (Purvis 2004; Maddison 2006).
A model system approach for macroevolution?
The combined weaknesses of the paleontological and phylogenetic approaches
greatly hamper macroevolutionary research, because there is currently no synthetic global overview of macroevolution in any large clade. However, a few
clades do at least begin to approach the ideal depicted in Fig. 15.1, and so might
usefully be developed as model systems for macroevolutionary research. The
planktonic foraminifera have the best species-level fossil record of any group
over the past 65 MY. These sexual unicellular marine zooplankton secrete
ornate calcareous shells (Fig. 15.5), and have morphologies that are specific to
biological species or closely related groups of cryptic species, of which around
50 are extant today (Hemleben et al. 1989; Norris 2000; Kucera & Darling 2002).
There is ongoing controversy about the meaning of species level categories both
in this group and in general, much of it semantic or arising from the desire to
apply hard and fast concepts to what is a loose natural category (Hey et al. 2003).
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
293 [278–300] 31.7.2008 11:11AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Figure 15.5 Fossil shell of a
planktonic foraminifera from
33.7 Mya (Hantkenina nanggulanensis).
Such shells occur in vast numbers in
stratified deep sea sediments and
provide an unrivalled opportunity for
examining the macroevolution of a
major taxonomic group by sampling
at the species level. Scale: about 1 mm
long.
The ‘species level’ in this context refers to clusters of related genotypes that
secrete shells of a recognizable unimodal type, sometimes changing through
anagenetic evolution but not by diversifying into more than one type, that can
often be traced for millions of years from their origin to either a branching event
(speciation) or their (often sudden) final extinction. The shells are deposited in
vast numbers in ocean sediments, where in favourable circumstances they can
accumulate continuously over many millions of years, thereby providing a
continuous record of the species’ existence from their origin to their extinction
(Parker et al. 1999). Many sites have now been cored in each of the world’s major
ocean basins and samples are also available from a wide range of reference
sections now exposed on land. The taxonomy of the group is well-developed and
mature (Olsson et al. 1999; Pearson et al. 2006), as is knowledge of their biostratigraphic distribution (Stewart & Pearson 2002). Furthermore, although the group
does not match the high levels of diversity of some other groups commonly used
in paleontological research, the fossil record is sufficiently complete that it is
even possible to sample large populations at will to focus on lineages or times of
particular interest, rather than rely on more or less haphazard finds of a few
specimens as is typical of most paleontological research.
293
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
294
294 [278–300] 31.7.2008 11:11AM
PURVIS ET AL.
The fossil record of planktonic foraminifera has been widely used for evolutionary research, including studies of morphometrics of individual lineages
(Malmgren & Kennett 1981; Malmgren et al. 1983), species survivorship
(Arnold 1982; Stanley et al. 1988; Pearson 1996; Parker & Arnold 1997), overall
size trends (Schmidt et al. 2004), diversity trends (Parker et al. 1999), tree shape
(Pearson 1998b; MacLeod 2001), shell chirality (Norris & Nishi 2001) and evolutionary rates (Prokoph et al. 2000; Allen et al. 2006). The group is ripe for a
thorough reinvestigation using the full range of phylogeny-based analytical
tools that have been developed recently.
For such a set of analyses to be successful, they must be based on a phylogeny
of evolutionary species-lineages rather than morphospecies.Morphospecies
(which are the standard units of taxonomy and biostratigraphy) are recognized
on the basis of similarity to the type specimen, whereas a species-lineage (or,
more simply, species) is a branch on a valid phylogenetic tree at the species level
(Fordham 1986; Pearson 1998b). The distinction is needed because a species may
evolve sufficiently that its members become recognized under different taxonomic names; morphospecies boundaries and even generic distinctions may in
principle be crossed without any new lineage being formed. Some recent macroevolutionary studies of the group (Prokoph et al. 2000; Allen et al. 2006; Doran
et al. 2006) have not made this distinction, so may have conflated taxonomic
turnover with anagenetic change.
Pearson (1993) constructed such a lineage phylogeny for Paleogene planktonic foraminifera. Although this phylogeny is somewhat out-of-date now, both
taxonomically and in terms of the timescale, it serves to illustrate the potential
that an updated and extended lineage phylogeny would have to shed light on
macroevolutionary dynamics (see also Pearson 1996; Pearson 1998b). Here, we
use it for a preliminary test a key assumption of the models at the heart of most
neontological macroevolution, namely that a species’ age has no bearing on its
chances of speciation or extinction (Kendall 1949; Hey 1992). Previous paleontological tests have used superspecific taxa, used morphospecies, been geographically restricted and/or been equivocal (Van Valen 1973; Parker & Arnold
1997; Alroy 1998; Pearson 1998b). We focused on only those internodes in the
phylogeny that terminated, either in speciation or extinction, before the end of
the period covered by the phylogeny. A logistic regression of fate (speciation or
extinction) against the length of the branch suggests that the balance between
speciation and extinction changed significantly with species age, with young
species being more likely to speciate and old species more likely to go extinct;
the balance point is at around 7 MY (Fig. 15.6). Further analyses could show
whether this change reflects age-dependency in speciation rates, extinction
rates or both, and could also test whether it is driven by differences in the
balance between speciation and extinction through geological time, or by
clade differences in rates. The system can also provide answers to a range of
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
295 [278–300] 31.7.2008 11:11AM
1.0
0.8
0.6
0.4
0.2
0.0
Probability species ends in speciation
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
0
5
10
15
Age of species (My)
20
Figure 15.6 The relative
chances of speciation or
extinction change as species’
age, for Paleogene planktonic
foraminifera. Circles:
branches in the phylogeny
that end in either speciation
(y = 1) or extinction (y = 0).
Thick line: fit from logistic
regression model, which is
highly significant (increase in
residual deviance on term
deletion = 10.502, 1 d.f,
p = 0.001). Thin grey line:
speciation and extinction are
equally likely, for any age of
species.
other questions which, although often addressed in the paleontological literature, cannot be tackled as unambiguously in many fossil groups (or in any
groups based solely on data from extant organisms). Does density-dependence
act by reducing speciation rates or raising extinction rates at high density? Does
the trend towards increasing size through time (Cope’s Rule: Arnold et al. 1995;
Webster & Purvis 2002) result from within-lineage changes alone, or do sizecorrelated speciation and extinction play a significant role (Alroy 2000)? And
have changes in environmental features such as climate or ocean chemistry
driven macroevolution? This hypothesis is long-standing (Stenseth & Maynard
Smith 1984) but remains controversial (Alroy et al. 2000; Gingerich 2006;
Jackson & Erwin 2006).
The model system approach has been very successful in many other areas of
biology, revealing many generalities that might never have emerged from less
intensive studies of a wider range of systems. We argue that the field of macroevolution would similarly benefit greatly from gaining detailed knowledge and
understanding of the most tractable systems, because of the light they may be
able to shed on some of the key processes that underpin diversity patterns in all
other groups. Much of the research based solely on extant diversity is reaching a
crucial stage, in which the very simple models of clade growth and character
change that gave the field its initial impetus are no longer providing useful new
directions for research. Demonstrating that clades reject these simple null
models is of some interest, but the problem is that there are many ways in
which the simple models might be made more complex, and data from extant
species alone provide little basis for choosing among them. Clades, such as
planktonic foraminifera, that have exceptional fossil records can provide the
295
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
296
296 [278–300] 31.7.2008 11:11AM
PURVIS ET AL.
impetus for the next generation of models, by highlighting which aspects of
complexity seem to be most needed.
Acknowledgements
We are very grateful to the editors for the invitation to the symposium, and to
J. Alroy, O.R.P. Bininda-Emonds and R.E. Ricklefs for very helpful comments on
the manuscript. This work was funded by the Natural Environment Research
Council (UK) and the Leverhulme Trust.
References
Allen, A. P., Gillooly, J. F., Savage, V. M. and
Brown, J. H. (2006) Kinetic effects of
temperature on rates of genetic divergence
and speciation. Proceedings of the National
Academy of Sciences of the United States of
America 103, 9130–9135.
Alroy, J. (1996) Constant extinction, constrained
diversification, and uncoordinated stasis
in North American mammals.
Palaeogeography, palaeoclimatology,
palaeoecology 127, 285–311.
Alroy, J. (1998) Equilibrial diversity dynamics in
North American mammals. In: Biodiversity
Dynamics: Turnover of Populations, Taxa and
Communities (ed. M. L. McKinney and
J. A. Drake), pp. 232–287. Columbia
University Press, New York.
Alroy, J. (1999) The fossil record of North
American mammals: evidence for a
Paleocene evolutionary radiation. Systematic
Biology 48, 107–118.
Alroy, J. (2000) Understanding the dynamics of
trends within evolving lineages. Paleobiology
26, 319–329.
Alroy, J., Koch, P. L. & Zachos, J. C. (2000). Global
climate change and North American
mammalian evolution. Paleobiology 26S,
259–288.
Alroy, J., Marshall, C. R., Bambach, R. K., et al.
(2001) Effects of sampling standardization
on estimates of Phanerozoic marine
diversification. Proceedings of the National
Academy of Sciences of the United States of
America 98, 6261–6266.
Arnold, A. J. (1982) Species survivorship in the
Cenozoic Globigerinida. Proceedings of the 3rd
North American Paleontology Convention 1,
9–12.
Arnold, A. J., Kelly, D. C. and Parker, W. C. (1995)
Causality and Cope’s Rule: evidence from
the planktonic Foraminifera. Journal of
Paleontology 69, 203–210.
Avise, J. C. and Walker, D. (1998) Pleistocene
phylogeographic effects on avian
populations and the speciation process.
Proceedings of the Royal Society of London Series
B-Biological Sciences 265, 457–463.
Avise, J. C., Walker, D. and Johns, G. C. (1998)
Speciation durations and Pleistocene effects
on vertebrate phylogeography. Proceedings of
the Royal Society B 265, 1707–1712.
Barraclough, T. G. and Nee, S. (2001)
Phylogenetics and speciation. Trends in
Ecology & Evolution 16, 391–399.
Barraclough, T. G. and Vogler, A. P. (2002) Recent
diversification rates in North American
tiger beetles estimated from a dated mtDNA
phylogenetic tree. Molecular Biology and
Evolution 19, 1706–1716.
Benton, M. J. and Donoghue, P. C. J. (2007)
Paleontological evidence to date the tree of
life. Molecular Biology and Evolution 24, 26–53.
Bininda-Emonds, O. R. P., Cardillo, M., Jones, K. E.,
et al. (2007) The delayed rise of present-day
mammals. Nature 446, 507–512.
Cifelli, R. L. and Gordon, C. L. (2007) Evolutionary
biology – re-crowning mammals. Nature
447, 918–920.
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
297 [278–300] 31.7.2008 11:11AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Crawley, M. J. (2002) Statistical Computing: An
Introduction to Data Analysis Using S-Plus. John
Wiley & Sons, New York and Chichester.
Cunningham, C.W., Omland, K.W. and
Oakley, T. H. (1998) Reconstructing ancestral
character states: a critical reappraisal. Trends
in Ecology & Evolution 13, 361–366.
Doran, N. A., Arnold, A. J., Parker, W. C. and
Huffer, F. W. (2006) Is extinction ageindependent? Palaios 21, 571–579.
Felsenstein, J. (1985) Phylogenies and the
comparative method. American Naturalist
125, 1–15.
Foote, M. (1996) Perspective: evolutionary
patterns in the fossil record. Evolution 50, 1–11.
Foote, M. (2000) Origination and extinction
components of taxonomic diversity:
Paleozoic and post-Paleozoic dynamics.
Paleobiology 26, 578–605.
Fordham, B. G. (1986) Miocene-Pleistocene
planktic Foraminifers from DSDP sites 208
and 77, and phylogeny and classification of
Cenozoic species. Evolutionary Monographs 6,
1–200.
Gill, F. B., Mostrom, A. M. and Mack, A. L. (1993)
Speciation in North American Chickadees:
I. Patterns of mtDNA divergence. Evolution
47, 195–212.
Gill, F. B. and Slikas, B. (1992) Patterns of
mitochondrial DNA divergence in North
American Crested Titmice. The Condor 94,
20–28.
Gingerich, P. D. (2006) Environment and
evolution through the Paleocene-Eocene
thermal maximum. Trends in Ecology &
Evolution 21, 246–253.
Harvey, P. H., May, R. M. and Nee, S. (1994)
Phylogenies without fossils. Evolution 48,
523–529.
Hemleben, C., Spindler, M. and Anderson, O. R.
(1989) Modern Planktonic Foraminifera.
Springer-Verlag, New York.
Hey, J. (1992) Using phylogenetic trees to study
speciation and extinction. Evolution 46,
627–640.
Hey, J., Waples, R. S., Arnold, M. L., Butlin, R. K.
and Harrison, R. G. (2003) Understanding
and confronting species uncertainty in
biology and conservation. Trends in Ecology &
Evolution 18, 597–603.
Jablonski, D. and Roy, K. (2003) Geographical
range and speciation in fossil and
living molluscs. Proceedings of the Royal Society
of London Series B-Biological Sciences 270,
401–406.
Jablonski, D., Roy, K. and Valentine, J. W. (2003)
Evolutionary macroecology and the fossil
record. In: Macroecology: Concepts and
Consequences (ed. T. M. Blackburn and
K. J. Gaston), pp. 368–390. Blackwell
Science, Oxford.
Jackson, J. B. C. and Erwin, D. H. (2006) What
can we learn about ecology and evolution
from the fossil record? Trends in Ecology &
Evolution 21, 322–328.
Johns, G. C. and Avise, J. C. (1998) A comparative
summary of genetic distances in te
vertebrates from the mitochondrial
cytochrome b gene. Molecular Biology and
Evolution 15, 1481–1490.
Johnson, K. P. and Sorenson, M. D. (1998)
Comparing molecular evolution in two
mitochondrial protein coding genes
(cytochrome b and ND2) in the dabbling
ducks (Tribe: Anatini). Molecular Phylogenetics
and Evolution 10, 82–94.
Kendall, D. G. (1949) Stochastic processes and
population growth. Journal of the Royal
Statistical Society, Series B 11, 230–264.
Kessler, L. G. and Avise, J. C. (1985) A comparative
description of mitochondrial DNA
differentiation in selected avian and other
vertebrate genera. Molecular Biology and
Evolution 2, 109–125.
Kirchner, J. W. and Weil, A. (2000) Delayed
biological recovery from extinctions
throughout the fossil record. Nature 404,
177–180.
Krajewski, C. and Fetzner, J. W., Jr. (1994)
Phylogeny of Ganes (Gruiformes: Gruidae)
297
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
298
298 [278–300] 31.7.2008 11:11AM
PURVIS ET AL.
based on cytochrome b DNA sequences. The
Auk 111, 351–365.
Kubo, T. and Iwasa, Y. (1995) Inferring the rates
of branching and extinction from molecular
phylogenies. Evolution 49, 694–704.
Kucera, M. and Darling, K. F. (2002) Cryptic
species of planktonic foraminifera: their
effect on palaeoceanographic
reconstructions. Philosophical Transactions of
the Royal Society of London A 360, 695–718.
Lovette, I. J. and Bermingham, E. (1999)
Explosive speciation in the New World
Dendroica warblers. Proceedings of the Royal
Society of London Series B-Biological Sciences 266,
1629–1636.
MacLeod, N. (2001) The role of phylogeny in
quantitative paleobiological data analysis.
Paleobiology 27, 226–240.
Maddison, W. P. (2006) Confounding
asymmetries in evolutionary diversification
and character change. Evolution 60,
1743–1746.
Malmgren, M. A., Berggren, W. A. and
Lohmann, G. P. (1983) Evidence for
punctuated gradualism in the Late Neogene
Globorotalia tumida lineage of planktonic
foraminifera. Paleobiology 9, 377–389.
Malmgren, M. A. and Kennett, J. P. (1981)
Phyletic gradualism in a late Cenozoic
planktonic foraminiferal lineage, DSDP Site
284, Southwest Pacific. Paleobiology 7,
230–240.
Nee, S. (1994) How populations persist. Nature
367, 123–124.
Nee, S. (2001) Inferring speciation rates from
phylogenies. Evolution 55, 661–668.
Nee, S. (2006) Birth-death models in
macroevolution. Annual Review in Ecology,
Evolution and Systematics 37, 1–17.
Nee, S., May, R. M. and Harvey, P. H. (1994) The
reconstructed evolutionary process.
Philosophical Transactions of the Royal Society of
London Series B 344, 305–311.
Nee, S., Mooers, A. Ø. and Harvey, P. H. (1992)
The tempo and mode of evolution revealed
from molecular phylogenies. Proceedings of
the National Academy of Sciences of the United
States of America 89, 8322–8326.
Norris, R. D. (2000) Pelagic species diversity,
biogeography, and evolution. Paleobiology
26, 236–258.
Norris, R. D. and Nishi, H. (2001) Evolutionary
trends in coiling of tropical planktic
foraminifera. Paleobiology 27, 327–347.
Nunn, G. B., Cooper, J., Jouventin, P.,
Robertson, C. R. J. and Robertson, G. G.
(1996) Evolutionary relationships among
extant albatrossses (Procellaroiformes:
Diomedeidae) established from complete
cytochrome b gene sequences. The Auk 113,
784–801.
Oakley, T. H. and Cunningham, C. W. (2000)
Independent contrasts succeed where
ancestor reconstruction fails in a known
bacteriophage phylogeny. Evolution 54,
397–405.
Olsson, R. K., Hemleben, C., Berggren, W. A. and
Huber, B. T. (eds) (1999) Atlas of Paleocene
Planktonic Foraminifera. Smithsonian
Contributions to Paleontology.
Smithsonian, New York.
Orme, C. D. L. (2002) Body size and
macroevolutionary patterns of speciesrichness. In Department of Biological Science,
Imperial College of Science, Technology and
Medicine, vol. PhD. University of London,
London.
Pagel, M. (1994) Detecting correlated evolution
on phylogenies: a general model for the
comparative analysis of discrete characters.
Proceedings of the Royal Society of London Series B
255, 37–45.
Paradis, E. (1997) Assessing temporal variations
in diversification rates from phylogenies:
estimates and hypothesis testing. Proceedings
of the Royal Society of London Series B-Biological
Sciences 264, 1141–1147.
Paradis, E. (1998) Detecting shifts in
diversification rates without fossils.
American Naturalist 152, 176–187.
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
299 [278–300] 31.7.2008 11:11AM
TEMPORAL PATTERNS IN DIVERSIFICATION RATES
Parker, W. C. and Arnold, A. J. (1997) Species
survivorship in the Cenozoic planktonic
foraminifera: a test of exponential and
Weibull models. Palaios 12, 3–11.
Parker, W. C., Feldman, A. and Arnold, A. J.
(1999) Paleobiogeographic patterns in the
morphologic diversification of the Neogene
planktonic foraminifera. Palaeogeography,
Palaeoclimatology, Palaeoecology 152, 1–14.
Pearson, P. N. (1993) A lineage phylogeny for the
Paleogene planktonic foraminifera.
Micropaleontology 39, 193–232.
Pearson, P. N. (1996) Cladogenetic, extinction
and survivorship patterns from a lineage
phylogeny: the Paleogene planktonic
foraminifera. Micropaleontology 42, 179–188.
Pearson, P. N. (1998a) Evolutionary concepts in
biostratigraphy. In: Unlocking the
Stratigraphical Record (ed. P. Doyle and
M. R. Bennett), pp. 123–144. John Wiley,
Chichester.
Pearson, P. N. (1998b) Speciation and extinction
asymmetries in paleontological
phylogenies: evidence for evolutionary
progress? Paleobiology 24, 305–335.
Pearson, P. N., Olsson, R. K., Huber, B. T.,
Berggren, W. A. and Hemleben, C. (eds)
(2006) Atlas of Eocene Planktonic Foraminifera.
Cushman Foundation, Cushman
Foundation Special Publications.
Penny, D. and Phillips, M. J. (2007) Evolutionary
biology – Mass survivals. Nature 446, 501–502.
Peters, S. E. and Foote, M. (2002) Determinants of
extinction in the fossil record. Nature 416,
420–424.
Pons, J., Barraclough, T. G., Gomez-Zurita, J., et al.
(2006) Sequence-based species delimitation
for the DNA taxonomy of undescribed
insects. Systematic Biology 55, 595–609.
Price, T., Gibbs, H. L., de Sousa, L. and
Richman, A. D. (1998) Different timing of
the adaptive radiations of North American
and Asian warblers. Proceedings of the Royal
Society of London Series B-Biological Sciences 265,
1969–1975.
Prokoph, A., Fowler, A. D. and Patterson, R. T.
(2000) Evidence for nonlinearity in a highresolution fossil record of long-term
evolution. Geology 28, 867–870.
Purvis, A. (2004) How do characters evolve?
Nature 432, doi:10.1038/nature03092.
Purvis, A., Nee, S. and Harvey, P. H. (1995)
Macroevolutionary inferences from primate
phylogeny. Proceedings of the Royal Society of
London Series B 260, 329–333.
Pybus, O. G. and Harvey, P. H. (2000) Testing
macro-evolutionary models using
incomplete molecular phylogenies.
Proceedings of the Royal Society of London Series B
267, 2267–2272.
Rabosky, D. L. (2006) Likelihood methods for
detecting temporal shifts in diversification
rates. Evolution 60, 1152–1164.
Raup, D. M. (1985) Mathematical models of
cladogenesis. Paleobiology 11, 42–52.
Schluter, D., Price, T., Mooers, A. Ø. and
Ludwig, D. (1997). Likelihood of ancestor
states in adaptive radiation. Evolution 51,
1699–1711.
Schmidt, D. N., Thierstein, H. R., Bollman, J. and
Schiebel, R. (2004) Abiotic forcing of
plankton evolution in the Cenozoic. Science
303, 207–210.
Sepkoski, J. J. (1996) Competition in
macroevolution: the double-wedge revisited.
In: Evolutionary Paleobiology (ed. D. Jablonski,
D. H. Erwin and J. H. Lipps), pp. 213–255.
University of Chicago Press, Chicago.
Sibley, C. G. and Ahlquist, J. E. (1990) Phylogeny
and Classification of Birds: A Case Study of
Molecular Evolution. Yale University Press,
New Haven, CT.
Simpson, G. G. (1953) The Major Features of Evolution.
Columbia University Press, New York.
Smith, A. B., Gale, A. S. and Monks, N. E. A. (2001)
Sea-level change and rock-record bias in the
Cretaceous: a problem for extinction and
biodiversity studies. Paleobiology 27, 241–253.
Stanley, S. M. (1979) Macroevolution.
W H Freeman, San Francisco.
299
//FS2/CUP/3-PAGINATION/SPDY/2-PROOFS/3B2/9780521883184C15.3D
300
300 [278–300] 31.7.2008 11:11AM
PURVIS ET AL.
Stanley, S. M., Wetmore, K. L. and Kennett, J. P.
(1988) Macroevolutionary differences
between the two major clades of Neogene
planktonic foraminifera. Paleobiology 14,
235–249.
Stenseth, N. C. and Maynard Smith, J. (1984)
Coevolution in ecosystems: red Queen
evolution or stasis? Evolution 38, 870–880.
Stewart, D. R. M. and Pearson, P. N. (2002)
Plankrange: a database of planktonic
foraminiferal ranges. http://palaeo.gly.bris.
ac.uk/Data/plankrange.html.
Tavaré, S., Marshall, C. R., Will, O., Soligo, C. and
Martin, R. D. (2002). Using the fossil record
to estimate the age of the last common
ancestor of extant primates. Nature 416,
726–729.
Turgeon, J., Stoks, R., Thum, R. A., Brown, J. M.
and McPeek, M. A. (2005) Simultaneous
Quaternary radiations of three damselfly
clades across the Holarctic. American
Naturalist 165, E78–E107.
Van Valen, L. (1973) A new evolutionary law.
Evolutionary Theory 1, 1–30.
Webster, A. J. and Purvis, A. (2002) Testing the
accuracy of methods for reconstructing
ancestral states of continuous characters.
Proceedings of the Royal Society of London Series B
214, 143–149.
Wible, J. R., Rougier, G. W., Novacek, M. J. and
Asher, R. J. (2007) Cretaceous eutherians
and Laurasian origin for placental
mammals near the K/T boundary. Nature
447, 1003–1006.
Wood, S. N. (2006) Generalized Additive Models: An
Introduction with R. Boca Chapman & Hall/
CRC, Raton, FL.
Yule, G. U. (1924) A mathematical theory of
evolution, based on the conclusions of
Dr. J. C. Willis F.R.S. Philosophical Transactions
of the Royal Society of London A 213, 21–87.
Zink, R. M. (1994) The geography of
mitochondrial DNA variation, popoulation
structure, hybridization, and species limits
in the Fox Sparrow (Passerella iliaca). Evolution
48, 96–111.
Zink, R. M. and Avise, J. C. (1990) Patterns of
mitochondrial DNA and allozyme evolution
in the avian genus Ammodramus. Systematic
Zoology 39, 148–161.
Zink, R. M. and Dittman, D. L. (1991) Evolution of
Brown Towhees: mitochondrial DNA
evidence. The Condor 93, 98–105.
Zink, R. M. and Dittman, D. L. (1993) Population
structure and gene flow in the Chipping
Sparrow and a hypothesis for evolution in
the genus Spizella. The Wilson Bulletin 105,
399–413.
Zink, R. M., Dittman, D. L., Klicka, J. and
Blackwell-Rago, R. C. (1999) Evolutionary
patterns of morphometrics, allozymes and
mitochondrial DNA in Thrashers (Genus
Toxostoma). The Auk 116, 1021–1038.
Zink, R. M., Dittman, D. L. and Rootes, W. L.
(1991a) Mitochondrial DNA variation and
the phylogeny of Zonotrichia. The Auk 108,
578–584.
Zink, R. M., Rootes, W. L. and Dittman, D. L.
(1991b) Mitochondrial DNA variation,
population structure, and evolution of the
Common Grackle (Quiscalus quiscula). The
Condor 93, 318–329.
Zink, R. M. and Slowinski, J. B. (1995) Evidence
from molecular systematics for decreased
avian diversification in the Pleistocene
epoch. Proceedings of the National Academy of
Sciences of the United States of America 92,
5832–5835.
Author Query for Chapter 15
AQ1
Please check the sentence “Cartoon illustrating…events.”